keviner1 / ffmef Goto Github PK

Official PyTorch implementation of our CVPRW2023 paper: Efficient Multi-exposure Image Fusion via Filter-dominated Fusion and Gradient-driven Unsupervised Learning. Paper link

Frameworks

FFMEF & GIFloss

Results

multi-exposure fusion

multi-focus fusion & visible-infrared fusion

We provide a simple training and testing process as follows:

Dependencies

• Python 3.8
• PyTorch 1.10.0+cu113

Train

The datasets samples are placed in images\dataset (including MEFB[1], MFIF[2], VIFB[3], and SICE[4]).

Multi-Exposure Image Fusion (MEF)

python train.py --config 1

Multi-Focus Image Fusion (MFF)

python train.py --config 2

Visible-Infrared Image Fusion (VIF)

python train.py --config 3

Then, the checkpoints and log file are saved in output.

Test

The pretrained models are placed in ckp.

MEF

python test.py --config 1 --ckp mef.pth

MFF

python test.py --config 2 --ckp mff.pth

VIF

python test.py --config 3 --ckp vif.pth

Finally, the fused results can be found in images\fused.

Citation

@inproceedings{zheng2023ffmef,
  title={Efficient Multi-exposure Image Fusion via Filter-dominated Fusion and Gradient-driven Unsupervised Learning},
  author={Zheng, Kaiwen and Huang, Jie and Yu, Hu and Zhao, Feng},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
  year={2023}
}

Reference

[1] Zhang X. Benchmarking and comparing multi-exposure image fusion algorithms[J]. Information Fusion, 2021, 74: 111-131.

[2] Zhang X. Deep learning-based multi-focus image fusion: A survey and a comparative study[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021.

[3] Zhang X, Ye P, Xiao G. VIFB: A visible and infrared image fusion benchmark[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. 2020: 104-105.

[4] Cai J, Gu S, Zhang L. Learning a deep single image contrast enhancer from multi-exposure images[J]. IEEE Transactions on Image Processing, 2018, 27(4): 2049-2062.